Uncertainty plays a fundamental role in structural engineering since it may a®ect both external excitations and structural parameters. In this study, the analysis of linear structures with slight variations of the structural parameters subjected to stochastic excitation is addressed. It is realistically assumed that su±cient data are available to model the external excitation as a Gaussian random process, while only fragmentary or incomplete information about the structural parameters are known. Under this assumption, a nonprobabilistic approach is pursued and the°uctuating properties are modeled as uncertain-but-bounded parameters via interval analysis. A method for evaluating the lower and upper bounds of the second-order statistics of the response is presented. The proposed procedure basically consists in combining random vibration theory with¯rst-order interval Taylor series expansion of the mean-value and covariance vectors of the response. After some algebra, the sets of¯rst-order ordinary di®erential equations ruling the nominal and¯rst-order sensitivity vectors of response statistics are derived. Once such equations are solved, the bounds of the mean-value and covariance vectors of the response can be evaluated by handy formulas.To validate the procedure, numerical results concerning two di®erent structures with uncertain-but-bounded sti®ness properties under seismic excitation are presented.